The scope limits are written into the tools rather than left for you to discover: the active search says outright that its results are registered domains, the market price reads Sedo and not Afternic or Dan, and the trend tool describes its output as a filtered sample rather than a registry feed or a valuation. The stateful tools stay inert until the account has memory enabled, and the strategy tool is storage — it holds text you wrote and the last result, and does not supply presets, pick domains or execute anything.
A domain research server over registration data — active, aged, expiring, deleted and newly registered names — with availability, WHOIS, DNS, pricing, trend and marketplace lookups alongside, plus a small stateful layer for monitors and preferences.
- Keyword search across the registered base: active over a database of ~240 million registered gTLD domains, aged over live names carrying years of history, nrds over recently registered domains, and market over names carrying marketplace listing data (active, aged, nrds, market).
- The acquisition window as its own search — expired for names in redemption or pending delete that are still held by their registrant, deleted for names that have finished the cycle and can be registered at standard cost today (expired, deleted).
- Availability answers that say which kind they are: a single domain's registrability with price, and a batch of up to 50 that separates available, registered, expiring, reserved and unknown rather than reading a failure as availability (available, bulk_available).
- TLD-level views: standard registration and renewal price, one prefix across the core TLDs, one keyword across many, and rankings and trends by registration volume over a period (price, tld_check, bulk_tld, tld_rank, tld_trends).
- Diligence on a single name — registrar, dates, status and nameservers; DNS records; everything else on the same nameserver; a backlink profile with domain rank and spam score; Google Safe Browsing status; and the Sedo asking price, Sedo only (whois, dns, ns_reverse, backlink_summary, safety, market_price).
- Brand protection by generation rather than guesswork: omission, transposition, keyboard-adjacent, insertion, repetition, hyphenation, vowel-swap, homoglyph, plural/singular and TLD-swap variants, each checked for whether it is registered (typosquat).
- Demand-side signals: search volume, CPC and competition for a keyword, hot and emerging keyword cohorts and prefix activity buckets, recent status changes on premium .com names, and unregistered short .ai names by pattern (keyword_data, keywords_trends, domain_changes, unregistered_ai).
- Stateful work that persists between sessions: monitors tracking WHOIS, DNS and page content with current and previous data flagged for change, plus your own preferences and strategy notes — all encrypted at rest with AES-256-GCM in per-user directories (monitor, preferences, strategy).
- Your own limits, readable: account tier, per-tool-group usage, rate limits, and the monitor and strategy quotas (usage).
Nothing to start: without DOMAINKITS_API_KEY the server runs on a guest quota, either locally over stdio through the npm package or by pointing a Streamable HTTP client at https://api.domainkits.com/v1/mcp. A key raises the tier. The stateful tools are different — monitors, preferences and strategy need a registered account with memory explicitly enabled, and they store nothing until it is. How many monitors you may keep and how often each tool group may be called follow the tier, which the usage tool reports.
One command — claude mcp add --transport http domainkits https://api.domainkits.com/v1/mcp